<html>
    <head>
      <base href="https://bugs.llvm.org/">
    </head>
    <body><span class="vcard"><a class="email" href="mailto:hahnjo@hahnjo.de" title="Jonas Hahnfeld <hahnjo@hahnjo.de>"> <span class="fn">Jonas Hahnfeld</span></a>
</span> changed
          <a class="bz_bug_link 
          bz_status_REOPENED "
   title="REOPENED - [AArch64] Incorrect RETURN_ADDRESS in tests"
   href="https://bugs.llvm.org/show_bug.cgi?id=36313">bug 36313</a>
          <br>
             <table border="1" cellspacing="0" cellpadding="8">
          <tr>
            <th>What</th>
            <th>Removed</th>
            <th>Added</th>
          </tr>

         <tr>
           <td style="text-align:right;">Resolution</td>
           <td>WONTFIX
           </td>
           <td>---
           </td>
         </tr>

         <tr>
           <td style="text-align:right;">Status</td>
           <td>RESOLVED
           </td>
           <td>REOPENED
           </td>
         </tr></table>
      <p>
        <div>
            <b><a class="bz_bug_link 
          bz_status_REOPENED "
   title="REOPENED - [AArch64] Incorrect RETURN_ADDRESS in tests"
   href="https://bugs.llvm.org/show_bug.cgi?id=36313#c17">Comment # 17</a>
              on <a class="bz_bug_link 
          bz_status_REOPENED "
   title="REOPENED - [AArch64] Incorrect RETURN_ADDRESS in tests"
   href="https://bugs.llvm.org/show_bug.cgi?id=36313">bug 36313</a>
              from <span class="vcard"><a class="email" href="mailto:hahnjo@hahnjo.de" title="Jonas Hahnfeld <hahnjo@hahnjo.de>"> <span class="fn">Jonas Hahnfeld</span></a>
</span></b>
        <pre>(In reply to Amara Emerson from <a href="show_bug.cgi?id=36313#c16">comment #16</a>)
<span class="quote">> Ok this isn't really a bug, this is a code generation difference between
> using SelectionDAG and FastISel. GlobalISel is only involved indirectly,
> since at -O0 GISel replaces FastISel and falls back to SelectionDAG, whereas
> before FastISel would handle the block with the inline ASM and not fall back
> to SDAG.

> FastISel has some specialized code in FastISel.cpp:fastEmitBranch() that
> will check if the successor block is the target of the unconditional branch,
> in which case it will not emit anything. SelectionDAG doesn't do that, so
> when GlobalISel gives up on the block, instead of trying FastISel, we fall
> back to SDAG and end up with the extra branch.</span >

Ok, this is the answer to "why do we have branches". But it doesn't explain why
GlobalISel falls back at all.

<span class="quote">> This is expected behavior in terms of falling back to SDAG, so I think the test > needs to be updated to handle this new codegen.</span >

As said yesterday I don't think it's possible to do this reliably. And I'd
argue that the generated code, while correct, doesn't make any sense.</pre>
        </div>
      </p>


      <hr>
      <span>You are receiving this mail because:</span>

      <ul>
          <li>You are on the CC list for the bug.</li>
      </ul>
    </body>
</html>